The ‘LexisNexis Publisher’ service offers large organizations tailored news by running custom predefined queries on the incoming news feeds. These feeds can either be manually curated by information specialists (‘editors’) at the customer side, which most users do, or the feeds can be pushed to the customers’ information management/distribution systems without human intervention. For the user of the services it is important that no relevant information is missed. Therefore, the filtering query needs to be adapted when there are important changes to the topic.
This can happen when new terminology becomes relevant for the topic (e.g. `wolf' for the topic `biodiversity'), there is a new stakeholder (e.g. the name of the new minister of economic affairs for the topic `industry and ICT') or new geographical names are relevant to the topic. (e.g. `Heumensoord' for the topic `refugees') The goal of the current work is to support users of news monitoring applications by providing them with suggestions for query modifications in order to retrieve more relevant news articles.
A demo application has been developed and used to collect feedback from 10 expert users of LexisNexis Publisher to determine the best method for generating term suggestions. In the application, a Boolean query can be entered that is used to search in Dutch newspapers. The found documents are shown in a result list and a list of query term suggestions (a pool of terms from all methods) is presented. Users were asked to judge the relevance of the returned terms on a 5-point scale, could update the search query (potentially with a suggested term) and retrieve a new result list.
The results of our user experiment show that with the best performing method, the user selected a term from the top-5 suggestion list for 13% of the topics, and judged at least one term as relevant for 25% of the topics. There is definitely room for improvement, but the users responded positively to the option of having a short list of suggestions that can be used to improve search queries. Below you can find a screenshot of the demo application and a video that shows the functionality of the demo application.